Briefly describe the difficulties in machine vision system design

machine vision systemThe difficulties in design include the following:

(1) Stability of lighting

Industrial vision applications are generally divided into four categories: positioning, measurement, detection and identification. Measurement has high requirements on the stability of illumination, because as long as the illumination changes by 10-20%, the measurement results may deviate by 1-2 pixels. This is not a software problem, it is a change in lighting, which causes a change in the position of the upper edge of the image. No matter how powerful the software is, it cannot solve the problem. The interference of ambient light must be eliminated from the perspective of system design, and at the same time, active lighting sources must be ensured. luminous stability. Of course, improving the resolution of hardware cameras is also a way to improve accuracy and resist environmental interference. For example, the spatial size of the previous camera counterpart was 10um per pixel, but after upgrading the resolution, it became 5um per pixel. The accuracy can be considered to be approximately doubled, and the interference to the environment is naturally enhanced.

(2) Inconsistency in workpiece position

Generally speaking, for measurement projects, whether it is offline detection or online detection, as long as it is a fully automated detection equipment, the first step is to find the target object to be measured. Every time the target to be measured appears in the shooting field of view, you must be able to accurately know where the target to be measured is. Even if you use some mechanical fixtures, etc., you cannot guarantee with high accuracy that the target to be measured will appear in the same position every time. Yes, this requires the use of the positioning function. If the positioning is inaccurate, the position of the measurement tool may not be accurate, and the measurement results may sometimes have large deviations.

(3) Calibration

Generally, the following calibrations are required for high-precision measurements: first, optical distortion calibration (if you are not using a software lens, calibration is generally required); second, projection distortion calibration, which is the image distortion correction represented by your installation position error. , the calibration of the three-object image space is to specifically calculate the size of the object space corresponding to each pixel.

However, the current calibration algorithms are all based on plane calibration. If the physics to be measured is not plane, the calibration will require some special algorithms to process, which cannot be solved by ordinary calibration algorithms.

In addition, for some calibrations, special calibration methods must be designed because the calibration board is not used. Therefore, the calibration may not be fully solved by the existing calibration algorithm in the software.

(4) The speed of movement of the object

If the object being measured is not stationary but in motion, then the effect of motion blur on image accuracy must be considered (blurred pixels = object movement speed * camera exposure time), which cannot be solved by software.

(5) Measurement accuracy of software

In measurement applications, the accuracy of software can only be considered in terms of 1/2-1/4 pixels, preferably 1/2, and cannot reach 1/10-1/30 pixel accuracy like positioning applications, because in measurement applications There are very few feature points that the software can extract from the image.